Demo clock: the simulated date is 2026-10-16. Data is synthetic.
Metrics
Computed from the app database when you load this page, at 2026-10-20T20:21:03+00:00 UTC. Day and week boundaries are real UTC dates, not the demo clock.
Conversation volume
| Today | Last 7 days | Total |
|---|
| 42 | 103 | 103 |
Containment
Conversations with at least one caller turn that were never handed to a human, divided by all conversations with at least one caller turn. This is a proxy: it does not know whether the caller got what they needed.
| Containment | Contained | With a caller turn |
|---|
| 76.5% |
13 | 17 |
Escalations by reason
| Reason | Handoffs |
|---|
| no_reliable_answer | 1 |
| safety:carbon_monoxide | 1 |
| safety:gas | 1 |
| safety:sparks_smoke_fire | 1 |
Failures
Total 0; last 24 hours 0. A failed turn gets a polite fallback reply and a handoff, not an error page.
| Kind | Count | Last seen (UTC) |
|---|
| No failures recorded. |
Latency per chat turn
Target (ours, not the client's; A9): chat p95 time to reply under 2 s.
| Turns | p50 | p95 | Target |
|---|
| 27 |
4329 ms |
10000 ms |
missed |
LLM spend
| Spent (USD) | Limit (USD) |
|---|
| Today (UTC) | 0.0177 | 2.00 |
| Total | 0.0200 | 10.00 |